Adaptive tracking control scheme for an autonomous underwater … · 2013. 12. 19. · Adaptive...
Transcript of Adaptive tracking control scheme for an autonomous underwater … · 2013. 12. 19. · Adaptive...
-
Adaptive tracking control scheme for an autonomous underwater vehicle
subject to a union of boundaries
Zool Hilmi Ismail1 & Mohd Bazli Mohd Mokhar
2
1 Robotics Research Group,
Centre for Artificial Intelligence and Robotics (CAIRO), Universiti Teknologi Malaysia,
Jalan Semarak, 54100, Kuala Lumpur, Malaysia.
2Faculty of Electrical Engineering, Universiti Teknologi Malaysia, 81310, Skudai, Malaysia.
[E-mail: 1
[email protected], 2 [email protected]]
Received ; revised
Present study presents a novel region boundary-based tracking control for an Autonomous Underwater
Vehicle (AUV). The control objective is to track a moving target formed by the union of all boundaries.
In this case, multiplicative potential energy function is used to unite all the boundaries and various shapes
can be created using this function. It is interesting to note that the AUV will be placed at a specific
position on the dynamic region boundaries.A Lyapunov-like function is presented for stability analysis of
the AUV. Simulation studies will be performed to illustrate the effectiveness of the proposed controller.
[Keywords: Multiplicative potential energy function, dynamic region boundary-based control; autonomous underwater vehicle]
Introduction
The current research studies of an Autonomous Underwater Vehicle (AUV) have focused on the development and
improvement of regulation and tracking control schemes. Breakthrough from these findings can be deployed for
numerous applications including industrial and military applications. Point stabilization is regarded as a problem of
commanding an AUV to a final target point with a desired orientation. Set-point control is one of the most
important topics in underwater vehicle navigation. However, when the desired target of the AUV will be a point, it
consumes a lot of energy due to propeller usually being active to counter the wave until the AUV arrived to its set-
point. Hence, an adaptive region tracking control is proposed by1. From their work, the target is defined as a moving
region, thus it would require less energy to perform docking or tracking. This is due to the controller only activated
when the AUV outside of the region rather than always pushing the AUV into the point target in conventional
method. The proposed technique not only applicable for an AUV but also can be utilized for a swarm of robots as
reported by2. While the robots are moving in a specific region, the formation can be rotated and/or scaled depending
on their mission. On the other hand,3 proposed a multiplicative energy function together with region control for a
-
swarm of robots. Within this control law, various shapes such as star shape and N-shape can be formed based on the
union of all the regions. In addition, a research work by4 combined the multiplicative potential energy function with
transformation mapping that enables the swarm robot to change their shape and size while moving. Recently, the
multiplicative potential energy function and adaptive boundary-based region was presented to control a single
AUV5. The vehicle has been able to navigate and locate itself to a desired position on a moving target using an edge-
based segmentation technique. Note that the respective moving target is created via the union of region boundaries.
In this paper, a dynamic boundary-based region tracking control approach and multiplicative potential energy
function are proposed for an AUV. The desired target is formed using the union of region boundaries and the AUV
will move in a specific direction alongside the moving region boundary as to monitor the pipeline thoroughly from
any degree. The region can be scaled up or down according to the state or size of the pipeline as depicted in Fig. 1.
Fig. 1 An illustration of dynamic boundary-based region control technique for tracking a pipeline
Materials and Methods
Modeling of an AUV
Kinematic Model
The relationship between inertial and body-fixed vehicle velocity can be described using the Jacobian
matrix in the following form
(1)
where and denote the position and the orientation of the
vehicle, respectively, expressed in the inertial-fixed frame. and are the transformation matrices
expressed in terms of the Euler angles. Linear and angular velocity vectors, and
, respectively, are described in terms of the body-fixed frame.
Dynamic Model
-
The investigation of dynamic equation of motion for an underwater vehicle has been previously reported6.
The underwater vehicle dynamic equation can be expressed in closed form as
(2)
where is the velocity state vector with respect to the body-fixed frame, is the inertia matrix
including the added mass term, represents the matrix of the Coriolis and centripetal forces including
the added mass term, denotes the hydrodynamic damping and lift force, is the restoring force
and is the vector of generalized forces acting on the vehicle. The dynamic equation in (2) preserves the
following properties6:
Property 1: The inertia matrix is symmetric and positive definite such that .
Property 2: is the skew-symmetric matrix such that .
Property 3: The hydrodynamic damping matrix is positive definite, i.e.: .
Property 4: The dynamic model as described in (2) is linear in a set of dynamic parameters and
can be written as
(3)
where is a known regression matrix; is the total number of physical parameters. It
is assumed that if the arguments of are bounded then is bounded.
Tracking Control Using Multiplicative Potential Energy Function
In region boundary-based control, the desired moving target is specified by at least two sub-regions
intersecting at the same point. Inner sub-region acts a repulsive region while the outer sub-region acts as
an attractive region. Regulation control concept that has been presented7 is extended for coordination
control of multiple AUVs8. A new proposed tracking control for an AUV subject to the union of
boundaries is formulated as follows:
-
First, a dynamic region of specific shape is defined and this can be viewed as a global objective of the
proposed control law. Global objective functions for outer sub-region and inner sub-region of AUV are
defined by the following inequalities:
(4)
where is the continuous first partial derivatives; is the time-varying reference point
inside the region. The following inequality function can be used for the inner sub-region
(5)
where the primary and secondary sub-regions share the same reference point, . Note that, (4) and (5)
are defined arbitrarily close to each other, such that
(6)
The corresponding potential energy function for the desired sub-region describes in (4) can be specified
as
(7)
where is positive scalar for an AUV. Similarly, the potential energy function for the inner sub-regions
in (5) can defined as follows
(8)
where is positive scalar. Differentiating (7) and (8) with respect to gives
(9)
(10)
Now, let (9) and (10) be represented as the primary region error and secondary region error
respectively in the following form
-
(11)
(12)
Next, an edge-based segmentation approach is utilized to ensure an AUV is placed at a desired position in
the target8. Potential energy function for a segmented boundary can defined as follows
(13)
where is a positive constant. Differentiating (13) with respect to gives
(14)
which leads to
(15)
To implement the multiplicative potential energy, let be the potential energy function associated with
region boundary
(16)
where is the number of desired boundaries. A multiplication method3 is adopted in this paper, thus the
total potential energy associated with the desired boundary in (16) is defined by
(17)
where is defined in (16). The desired boundary produced from this multiplicative of the potential energy
is the union of all the boundaries that is . Note that has a minimum
value of zero when is within any of the desired boundaries. Equation (17) expresses that the potential
-
energy is at the minimum value (zero) at the desired target. This potential function will ensure the AUV
move toward the overall region produced by union of all the boundaries , , ..., . This function
is useful when the AUV need to adapt the moving boundary, depending on the situation and environment
such as avoiding obstacle on its path or capturing images at some different angles.
Partial differentiating the total potential energy function described by (17) with respect to leads to
(18
)
where the product rule is used to obtain the derivatives of products of two or more functions. When the
AUV are outside the desired boundary, the control force described by (18) is activated to attract the
AUV toward the desired boundary. When the AUV is inside the desired boundary, then the control force
is zero or . Next, a vector that is useful is defined
(19)
where is the inverse of the Jacobian matrix, and is a positive constant. The error term is given in
(18). Based on the structure of (18) and (19) and the subsequent stability analysis, a filtered tracking error
vector for an underwater vehicle is defined as
(20)
In general, the development of the open-loop error system for can be obtained by pre-multiplying the
inertia matrix with the time derivative of to yield
(21)
where
(22)
-
and the derivative of in (22) is given as
(23)
Based on the error system development and the subsequent stability analysis, the proposed adaptive
control law for AUV is
(24)
where is positive constant matrix. The estimated parameters are updated using the following update
law
(25)
where is a symmetric positive definite. Substituting (24) into (21) produces a closed-loop dynamic for
as follows
(26)
where denotes the parameter estimation error. Next, the following non-negative function is
introduced to analyze the stability of the proposed control law
(27)
Differentiating with respect to time and using the update law (25) yields
(28
)
Utilizing equation (19), (20), a closed-loop dynamic (26) and cancelling the common terms leads to
(29)
where Property 3 is used. Now, a new theorem can be stated as follows:
-
Theorem: Given a closed-loop of AUV in (26), the proposed adaptive control law (24) and the update
parameter laws (25) guarantees the convergence of into a dynamic region boundary in the sense that
and , as .
Proof: See previous work by Ismail et al.5 for proof.
Remark: The proposed dynamic control concept can be extended to the case of rotating and scaling
region boundary. In this case, continuous first partial derivatives in (4) needs to be exploited such that
; is rotational matrix given as
(30)
where is an angle of rotation. Generally, the rotation matrix R need not be specified for orientation
function. Meanwhile is a time-varying and nonsingularscaling factor that is defined as
(31)
where is the scaling matrix of and is the scaling matrix of . The scaling of the orientation of
AUV is not required in general, so can be set as an identity matrix. Thus, the scaling matrix is given
by
(32)
where and are scaling factors.
Results and Discussion
Simulation Results
In this section, two simulation studies are carried out to assess the efficacy of the proposed dynamic
region boundary-based control law for an underwater vehicle. The vehicle is required to track a pipeline
in this simulation. An ODIN with full 6-DOF9 is chosen as autonomous underwater vehicle model for
numerical simulation. In simulation, the following inequality functions are defined for a boundary of a
spherical region
-
(33)
(34)
while the subsequent inequality functions are defined for a boundary of a ellipsoid region
(35)
(36)
Equations (33) and (34) represent the outer sub-regions and (35) and (36) represent the inner sub-regions.
is a time-varying tolerance vector. An underwater vehicle isrequired to track a straight-line
trajectory with green and magenta (cross-section lines) trajectory is the horizontal basis position
initialized at m. The green and magenta designate spherical and ellipsoid regions,
respectively. The solid blue lines represent the position of an AUV at various time instances.
In the first simulation, the vehicle is initialized to the position m and it is
required to converges into the bottom of a moving scalable spherical region, while in the second
simulation, the vehicle is initialized to the position m and it is required to
converges the top of a moving scalable ellipsoid region. The scaling matrix presented in (32) is used in
both simulations. The orientation is kept constant with the allowable errors, denoted by , and ,
are set to rad, and the initial values are rad. To ensure the vehicle moves towards
its desired position on a moving boundary, the following inequality function is defined
(37)
where the vector is the circumradius for an AUV and is a tolerance
vector. The control gains are set to the following:
; ; ; ;
-
As can be seen from Fig. 2 to Fig. 5, an AUV initially converges into the desired position on a boundary
line at various time instances.
Fig. 2. A 3-dimension view of simulation study 1; marks the initial position of an AUV.
Fig. 3. Planar trajectories of an AUV for simulation study 2that are illustrated in (a) XZ-plane and (b)
YZ-plane.
Fig. 4. A 3-dimension view of simulation study 2; marks the initial position of an AUV.
Fig. 5. Planar trajectories of an AUV for simulation study 2 that are illustrated in (a) XZ-plane and (b)
YZ-plane.
Conclusion
In this paper, a new dynamic region boundary-based method has been proposed for an autonomous
underwater vehicle. This control technique enables an AUV to perform a specific underwater tracking
task. It has been shown that AUV is able to track a desired moving boundary produced by the union of
two or more boundaries. Moreover, an AUV navigates into a specific location on the boundary lines or
surfaces while the target itself is moving. The Lyapunov-like function is used to analyze the stability of
the controller. Simulation results have been presented to demonstrate the performance of the proposed
tracking controller.
Acknowledgments
This work was supported in part by the Ministry of Higher Education, Malaysia and Universiti Teknologi
Malaysia under Grant no. R.J130000.7823.4F101.
References
1. Li, X., Hou, S. P., and Cheah, C. C., Adaptive Region Tracking Control for Autonomous
Underwater Vehicle, in Proceedings of the 11th Int. Conf. on Control, Automation, Robotics and
Vision, Singapore, 2010, pp. 2129-2134.
-
2. Hou, S. P., Cheah, C. C., and Slotine, J. J. E., Dynamic Region Following Formation Control for
a Swarm of Robots, in Proceedings of the 2009 IEEE international conference on Robotics and
Automation, 2009, pp. 1528-1533.
3. Hou, S. P., and Cheah, C. C., Multiplicative Potential Energy Function for Swarm Control, in
Proceedings of the 2009 IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, St. Louis, USA,
2009, pp. 4363-4368.
4. Hou, S. P., and Cheah, C. C., Dynamic Compound Shape Control of Robot Swarm, IET Control
Theory and Applications , vol. 6, no. 3, pp. 454-460, Feb. 2012.
5. Ismail, Z. H., Mokhar, B. M., and Dunnigan, M. W., Tracking Control for an Autonomous
Underwater Vehicle Based on Multiplicative Potential Energy Function, in Proceedings of the
OCEANS 2012 MTS/IEEE , Yeosu, Republic of Korea, 2012.
6. Fossen, T. I., Guidance and Control of Ocean Vehicles, 1st ed. New York: John Wiley and Sons,
1994.
7. Ismail, Z. H., and Dunnigan, M. W., A Region Boundary-Based Control Scheme for an
Autonomous Underwater Vehicle, Ocean Engineering, vol. 38, no. 17-18, pp. 2270-2280, 2011.
8. Ismail, Z. H., and Dunnigan, M. W., Geometric Formation-Based Region Boundary Control
Scheme for Multiple Autonomous Underwater Vehicles, in Proceedings of the Int. Conf. on
Electrical, Control and Computer Engineering, Pahang, Malaysia, 2011, pp. 491-496.
9. Choi, S., Yuh, J., and Keevil, N., Design of Omni-Directional Underwater Robotic Vehicle, in
OCEANS ’93, Engineering in Harmony with Ocean, Proc., vol. 1, 1993, pp. I192-I197.
Fig. 1 An illustration of dynamic boundary-based region control technique for tracking a pipeline
Fig. 2. A 3-dimension view of simulation study 1; x_1marks the initial position of an AUV.
(a)
-
(b)
Fig. 3. Planar trajectories of an AUV for simulation study 2that are illustrated in (a) XZ-plane and (b)
YZ-plane.
Fig. 4. A 3-dimension view of simulation study 2; marks the initial position of an AUV.
(a)
(b)
Fig. 5. Planar trajectories of an AUV for simulation study 2 that are illustrated in (a) XZ-plane and (b)
YZ-plane.